663580 - Determine if private browsing causes HTTP cache invalidation/flushing

Status: RESOLVED INVALID

(Core :: Networking: Cache, defect)

Description

[Jason Duell]

We're getting reports (see bug 651011 comment 0 and 7) of lots of cache I/O when leaving Private Browsing mode.   We should look into whether/why this is happening.  Not clear if it's cache invalidation or some other thing that might cause lots of eviction.

Comment 1

[Josh Matthews [:jdm]]

Entering Private Browsing clears the cache.

http://mxr.mozilla.org/mozilla-central/source/browser/components/privatebrowsing/src/nsPrivateBrowsingService.js?force=1#571

588       // NOTE: there is no way to clear just that domain, so we clear out
589       //       everything)
590       try {
591         cs.evictEntries(Ci.nsICache.STORE_ANYWHERE);
592       } catch (ex) {
593         Cu.reportError("Exception thrown while clearing the cache: " +
594           ex.toString());
595       }

Comment 2

[Josh Matthews [:jdm]]

No, I'm sorry, that's incorrect. The only places that actually call removeDataFromDomain are about:permissions and http://mxr.mozilla.org/mozilla-central/source/browser/components/places/content/controller.js#257, so that's potentially a red herring.

Comment 3

[Josh Matthews [:jdm]]

Aha, I knew I wasn't making things up: http://mxr.mozilla.org/mozilla-central/source/netwerk/cache/nsCacheService.cpp#2515

That's called every time private browsing is entered or exited.

Comment 4

[Jason Duell]

Will get fixed in bug 666059

Comment 5

[Michal Novotny [:michal]]

When leaving PB mode we don't delete the disk cache as mentioned in description of bug #666059. At least not internally in cache service. We disable disk (and offline) cache when entering PB mode and enable it again when leaving the PB mode.

Comment #1 points to a JS code that evicts all entries and AFAICS this has nothing to do with leaving PB. But I'm having hard time to find out when exactly is this code called. I've found these references:

http://mxr.mozilla.org/mozilla-central/source/browser/components/places/content/placesOverlay.xul#109
http://mxr.mozilla.org/mozilla-central/source/browser/components/preferences/aboutPermissions.js#268

Comment 6

[Josh Matthews [:jdm]]

As pointed out in comment 2, comment 1 is bogus. Comment 3 is where the facts are.

Comment 7

[Michal Novotny [:michal]]

The line numbers have changed. Anyway if you mean nsCacheService::OnEnterExitPrivateBrowsing() then this bug is invalid since as I wrote in comment #5 we don't delete the disk cache.

If there is sometimes really lot of cache I/O we need a STR since I can't reproduce it.

Comment 8

[Patrick McManus [:mcmanus]]

invalid based on comment 7.. if it should be reopened please do so